@article{oai:nagoya.repo.nii.ac.jp:00021577, author = {松崎, 哲朗 and 小田, 昌宏 and 北坂, 孝幸 and 林, 雄一郎 and 三澤, 一成 and 森, 健策 and MATSUZAKI, Tetsuro and ODA, Masahiro and KITASAKA, Takayuki and HAYASHI, Yuichiro and MISAWA, Kazunari and MORI, Kensaku}, issue = {410}, journal = {電子情報通信学会技術研究報告. MI, 医用画像}, month = {Jan}, note = {腹部血管は複雑な分岐構造を有するため,安全な外科手術を行う上でその構造把握は重要である.そこで,本稿では腹部動脈および肝門脈系の分岐構造を解析する手法を提案する.血管は木構造として表現され,機械学習を用いた手法によりそれぞれの枝が各血管である尤度を算出する.血管のとりうる分岐パターンはグラフとして表現され,そのグラフの各辺は尤度を基に重み付けられる.このグラフの全域木は1つの分岐パターンを表現しており,重みを最大化するものを求めることにより分岐パターンの判定を行うことができる.求めた分岐パターンを基に木構造の各枝に解剖学的名称を対応付ける.腹部CT像50例で実験を行ったところ,80.8%の分岐パターンを正しく判定できた., Since abdominal blood vessels have complicated branching structures, understanding them is important to perform abdominal surgeries. In this paper, we propose a method for automated analysis of branching structures of abdominal arteries and hepatic portal system. A blood vessel region is expressed as a tree structure. Likelihoods of candidate anatomical names for each branch in the tree structure is computed by utilizing a machine learning-based method. Possible branching patterns are expressed as a graph whose edges are assigned weights based on the likelihoods of the branches in the tree structure. The directed spanning tree of the graph represents a branching pattern. The optimum branching pattern is obtained by computing the directed maximum spanning tree. Each branch in the tree structure is labeled anatomical names based on the branching pattern. In an experiment using 50 cases of abdominal CT volumes, 80.8% of branching patterns are obtained correctly., IEICE Technical Report;MI2013-59}, pages = {19--24}, title = {分岐パターン解析による腹部動脈および肝門脈系に対する解剖学的名称の自動対応付け}, volume = {113}, year = {2014} }